In conventional media systems, the interactions of a user with media content may be tracked. For example, a content provider may identify the programs a user has watched and recommend similar programs to the user. Likewise, the content provider may provide targeted advertisements (e.g., advertisements targeted to a user based on the interests, demographics, etc., of the user) to the user. While such “user profiling” may aid content providers in providing content customized for the users, such profiling has its limitations. Specifically, the user profile may contain incorrect data, and thus provide incorrect content recommendations. For example, the content provider may have erroneously associated the viewing of a program with the wrong user (e.g., a different user may have watched content while a device was associated with the user), the user profile may contain out-of-date information (e.g., the tastes of the user may have changed), and/or the criteria used to classify programs, and/or the criteria that a program meets in the user profile may have changed (e.g., the genre classification of a program a user views has changed from “horror” to “comedy”).